Токены ИИ взвинчивают корпоративные облачные счета — ZDNet фиксирует тренд

Corporate spending on AI tokens is turning into a new budget line item that nobody knows how to control. ZDNet analysts are registering an alarming trend: the cost curve is repeating the early history of the cloud — explosive growth without clear optimization tools and without an answer to the question of what business gets in return.

Déjà vu from 2012

In the early days of cloud, IT directors received bills for EC2, S3, and RDS that grew faster than any forecast. Back then, nobody knew how to calculate spending per unit of business value. It took years, dozens of FinOps tools, and entire cloud cost management departments to bring these expenses under control.

With AI tokens, the story is repeating itself, but at an accelerated pace. Developers are connecting GPT-4o, Claude Opus, and proprietary fine-tuned models to product systems. Each agentic workflow with tool calls, RAG search, and long context can spend from 20,000 to 100,000 tokens per single transaction.

Multiply that by thousands of users per day — and you get a bill that was never in any budget and keeps growing every quarter.

"AI tokens will remind many corporate clients of early cloud pricing,"

ZDNet notes.

Unsolved problem: how to measure value

The main issue ZDNet identifies is not the spending itself, but that companies don't know how to measure the value that AI tokens create. In the cloud world everything was relatively transparent: an instance runs a task, the task has a cost in man-hours, the difference is savings. With AI, the scheme breaks down at every step: An AI assistant speeds up email writing, but by how much exactly — nobody measures this systematically An agent automates a process, but the quality of its work is subjective and inconsistent A chatbot relieves support, but satisfaction metrics on complex cases drop A code assistant reduces development time, but technical debt doesn't disappear * A RAG system improves answer accuracy, but benchmarks depend on the dataset and specific task Unlike CPU hours or gigabytes of traffic, tokens don't have an obvious correspondence to business results.

A CFO can't say: "For this million tokens we got this specific measurable value." Until this problem is solved, AI budgets will be approved on faith, not on data.

How the market responds

Some large corporate clients of AWS, Azure, and Google Cloud are already seeing AI spending grow 3-10x year over year. Providers respond by releasing token consumption monitoring tools — but so far they mainly show numbers rather than help optimize spending. A new specialization is forming — AI FinOps, whose task is to manage the cost of LLM inference in production.

Among the first tactical tools: Prompt caching for repeated requests Task routing to cheaper models depending on complexity Limiting context depth and number of steps in agent chains Batching requests instead of single real-time calls * Regular audits of unused or poorly performing AI integrations But this is tactic, not strategy. The industry still has no answer to the question "how many tokens should we spend to get a specific business result."

What this means

Corporate AI is falling into the same trap that the cloud fell into in 2012: the technology is widely adopted, expenses are growing rapidly, ROI is hard to measure. Companies that have already deployed LLMs in production should invest in cost visibility tools and build value metrics right now — otherwise, the conversation with the CFO will turn into an awkward exam that nobody prepared for.